Your place or mine?
I started typing a comment to previous, but then decided that it was too long to bury in a blog comment. By the way, the reason Kurt couldn't comment directly is blog spam. I just couldn't keep up with the comment spams, and blosxom doesn't offer a spamfilter.
Anyway, that's just an aside -- what I meant was, if people start writing a usability report on the current version of Krita with lots of good ideas on how to improve, say, the toolbar, won't they be disappointed to learn that we were going to implement a new toolbox anyway?
On a deeper level, the reason it's hard to involve usability experts in an early stage is that much development in those stages is exploratory, not coding to reach an already conceived-of state, but just futzing around, adding a bit of code here & there. Getting a feel for the problem domain, even. I mean, for a long time the only book on computer graphics I had when coding for Krita was Java2D Graphics by Jonathan Knudsen, because my company had a copy on its shelves. I knew nothing about graphics, and I'm still not an expert.
And Krita had other problems: three different views of what an image editor should be intermingled, with layer upon layer of architectural and accidental residue.
Just getting most things to work within the existing framework took me and the rest of Krita hackers about a year and a half. Take a look at my first baby steps, in January 2004...
Most things work now, many things have been re-done, in some cases for the sixth time in as many years, and finally we've got an application that the original authors and the subsequent maintainers might recognize as something that comes close to what they had intended KImageShop, Krayon, Kandinsky or Krita to be. My own idea of what a paint app should be is quite different, but these are baby steps and I'm learning all the way.
Anyway, just before Kurt published his comment, I'd registered Krita with OpenUsability. Let's wait and see if there are people willing to be consulted about questions like "what should happen when a user applies a rotation to a layer that has a selection active" or "what is better -- having a button on-screen to execute and commit a crop command, or execute and commit the crop on changing to another tool". I do hope so, because not every problem can be solved by doing what the leading commercial application does.